“ “RReessttrruuccttuurriinnggiinnpprriivvaattiisseeddffiirrmmss::aaSSttaattiiss a apppprrooaacchh”” A Addeellaaiiddee FFiigguueeiirreeddoo F Feerrnnaannddaa FFiigguueeiirreeddoo N Naattáálliiaa PP.. MMoonntteeiirroo O Odddd RRuunnee SSttrraauummee NIPE WP 16/ 2010
“
“
R
R
e
e
s
s
t
t
r
r
u
u
c
c
t
t
u
u
r
r
i
i
n
n
g
g
i
i
n
n
p
p
r
r
i
i
v
v
a
a
t
t
i
i
s
s
e
e
d
d
f
f
i
i
r
r
m
m
s
s
:
:
a
a
S
S
t
t
a
a
t
t
i
i
s
s
a
a
p
p
p
p
r
r
o
o
a
a
c
c
h
h
”
”
A
A
d
d
e
e
l
l
a
a
i
i
d
d
e
e
F
F
i
i
g
g
u
u
e
e
i
i
r
r
e
e
d
d
o
o
F
F
e
e
r
r
n
n
a
a
n
n
d
d
a
a
F
F
i
i
g
g
u
u
e
e
i
i
r
r
e
e
d
d
o
o
N
N
a
a
t
t
á
á
l
l
i
i
a
a
P
P
.
.
M
M
o
o
n
n
t
t
e
e
i
i
r
r
o
o
O
O
d
d
d
d
R
R
u
u
n
n
e
e
S
S
t
t
r
r
a
a
u
u
m
m
e
e
N
N
I
I
P
P
E
E
**W
W
P
P
1
1
6
6
/
/
2
2
0
0
1
1
0
0
URL: http://www.eeg.uminho.pt/economia/nipe *NIPE – Núcleo de Investigação em Políticas Económicas – is supported by the Portuguese Foundation for Science and Technology through the Programa Operacional Ciência, Teconologia e Inovação (POCI 2010) of the
Restructuring in privatised firms:
a Statis approach
Adelaide Figueiredo∗ Fernanda Figueiredo† Nat´alia P. Monteiro‡ Odd Rune Straume§
Abstract
We analyse the dynamics and evolution of the corporate restructuring process in the Portuguese banking sector, where 10 banks were privatised during the period 1989-1996. We apply a novel methodological approach in this context, using a multi-dimensional measure of restructuring that links product and labour market variables. We find evidence of considerable heterogeneity in the restructuring process, where firms adjust at different speeds and intensities. We also find that the wage level is by far the firm attribute that changed more, which is shown to reflect substantial changes in terms of composition, and not size, of the workforce. Our empirical evidence also suggests that privatisation is associated with a higher level of rent sharing.
Keywords: Statis, privatisation, Portuguese banking JEL Classifications: C49, D21, L33.
∗
Faculty of Economics and LIAAD/INESC-Porto, University of Porto; Rua Dr. Roberto Frias, 4200-464 Porto, Portugal. E-mail: adelaide@fep.up.pt
†
Faculty of Economics, University of Porto and CEAUL; Rua Dr. Roberto Frias, 4200-464 Porto, Portugal. E-mail: otilia@fep.up.pt
‡
Corresponding author. Department of Economics and NIPE, University of Minho, Campus de Gualtar, 4710-057 Braga, Portugal. E-mail: n.monteiro@eeg.uminho.pt
§
Department of Economics and NIPE, University of Minho, Campus de Gualtar, 4710-057 Braga, Portugal; and University of Bergen (Economics), Norway. E-mail: o.r.straume@eeg.uminho.pt
1
Introduction
One of the most important structural economic reforms taking place during the last three
decades is the privatisation of state-owned enterprises in a wide range of industries all
over the world. In the present paper we introduce a new methodological approach to
study empirically the dynamics and evolution of the corporate restructuring process due
to privatisation, using a multidimensional measure of the type and extent of restructuring,
where we link product and labour market variables.
There is now a voluminous empirical literature evaluating the effects of various
pri-vatisation reforms. This literature can broadly be classified into two different categories.
First, there is a huge body of literature studying the effects of privatisation on various
measures of performance, firm-specific or economy-wide. Then, there is a much smaller
but rapidly growing body of literature on the effects of privatisation on labour market
outcomes, foremostly wages and employment.
The main bulk of the literature in the former category is summarised in four relatively
recent surveys. Megginson and Netter (2001) provide an extensive review of privatisation
effects worldwide, but focussing mainly on Western European (non-transition) countries,
while Megginson and Sutter (2006) survey empirical studies of privatisation effects in
de-veloping countries. On the other hand, Djankov and Murrell (2002) and Estrin et al.
(2009) concentrate on privatisation effects in (post-communist) transition economies and
China. Although the results vary considerably across countries and industries, and
ac-cording to different characteristics of the privatisation process, it is a widespread finding
that privatisation leads to improved performance, measured either by firm-specific
indica-tors like profits, productivity or value added, or by economy-wide indicaindica-tors like economic
growth or total factor productivity.
Apart from the effect of privatisation on the total employment level of the firm, which
is sometimes used as an extra indicator of firm performance, the studies reviewed in the
above-mentioned surveys are largely silent on the effect of privatisation on labour market
privatisation, the most commonly studied causal relationship is that between privatisation
and wages. Although results are not unanimous, the broad picture painted from these
studies is that privatisation generally leads to higher wages for workers in the privatised
firms. This finding is reported by Bishop and Kay (1988) and Parker and Martin (1996)
for UK data, La Porta and Lopez-de-Silanes (1999) for Mexican data, Brainerd (2002)
for Russian data, Ho et al. (2002) for Chinese data, Brown et al. (2006) for Ukrainian
data and Monteiro (2010) for Portuguese data.1 On the other hand, employment effects
tend to be more ambiguous and are often found to be negligible (Brown et al., 2010)
or negative (Haskel and Szymanski, 1993; La Porta and Lopez-de-Silanes, 1999).2 One
possible explanation for these observed labour market effects is that privatisation tends
to increase labour productivity, which is also consistent with the often found result that
privatisation enhances performance.
Studies that try to relate both product and labour market effects of privatisation are
scarce, as most of the empirical literature on privatisation effects of firm performance
builds on data obtained at firm level from company records. This type of data is usually
very crude in terms of labour outcomes. On the other hand, studies that look at effects
of ownership changes on wages – using data at individual level – usually have little or no
information at firm level.3 Thus, most privatisation studies are partial in the sense that
they focus on a limited number of variables, related to either the product or the labour
market. However, the corporate restructuring that takes place as a result of a transfer of
ownership, from public to private, is arguably a highly dynamic, multi-dimensional and
complex process. Motivated by this premise, our study departs from the existing empirical
1A decomposition of wage effects can often reveal more ambiguous results. For example, Brown et al. (2006) find that wages drop in privatised firms that are worker-controlled, while Monteiro (2010) find short-term wage losses for workers in privatised firms (that are more than compensated for by the long-term wage gains).
2
There are also several studies focussing on the effect of privatisation on other labour market outcomes (besides average wages and total employment), such as wage discrimination (Liu et al., 2000; Peoples and Talley, 2001), wage inequality (Brainerd, 1998; Ho et al., 2002), returns to seniority or experience (Munich et al., 2005) and labour force composition (Chong and Leon, 2009).
3
A related strand of the literature that relates product and labour markets is the literature on rent sharing between firm owners and their workers. However, this literature is largely silent on the effect of different types of firm ownership. A recent exception is Monteiro et al. (2010) who find that private firm ownership is associated with a higher level of rent sharing.
privatisation literature in two important ways.
First, we do not confine our analysis of the effects of privatisation to separate analyses
with multiple measures of firm performance or labour market outcomes. Instead, we
build a multidimensional measure of restructuring that simultaneously takes into account
various aspects of both the product and the labour market. More precisely, for each point
in time, we define an object that summarizes different aspects (related to labour and
product markets) of the privatised firms. Our measure of corporate restructuring is then
given by the RV coefficient that measures the distance between two objects defined at
two points in time. Thus, compared to previous literature, our multidimensional measure
constitutes a more complete summary of the restructuring process. Furthermore, as a
separate contribution to the literature, the creation of this multidimensional measure also
allows us to trace changes in the composition of privatised firms’ workforce, an aspect that
has been largely neglected in the privatisation literature.
Second, our methodological approach allows us to take directly into account the
dy-namics and evolution of the restructuring process following privatisation. The Statis
methodology used in the present paper is a three-way exploratory technique based on
Principal Components Analysis which includes two methods: the Statis and dual Statis.
Both methods allow us not only to build an object and derive the corresponding RV
co-efficient, but also to decompose the deviations between two data tables across firms or
variables. This means that we are able to assess and compare – at each point in time
– the extent of restructuring in each firm and the magnitude of change in each variable
measured. This approach is novel to the literature and we believe that it is particularly
well suited to study the restructuring process of corporate privatisation.4 Our approach
also contrasts sharply with previous approaches. In particular, we do not aim to establish
a causal relationship between privatisation and different measures of firm performance or
4
As noted by Villalonga (2000, p. 51):
”privatization is by definition a change, and needs to be addressed dynamically by looking at a given firm’s evolution and transition between its private and public stages within a given firm.”
labour market outcomes. Instead, we aim to capture the dynamics and understand the
evolution and interrelation among the different dimensions of the restructuring process.
Our approach (to be described in detail later) is applied to the privatisation, and
subsequent corporate restructuring, in the Portuguese banking industry between 1989
and 1997. We select this industry for different reasons. First, until the mid-1990s, the
Portuguese privatisation programme was asymmetric and sectorally biased. Its major
incidence – in terms of number of firms or total revenues generated – was in banking.
The privatisation process comprised eleven companies, which accounted for more than
83% of banking employment in 1985 and raised 3,3 billions of euros, which constituted
almost half of the total sales of state enterprises in Portugal until the second quarter of
1995. Moreover, in contrast with some other economic sectors, where privatisation is less
advanced and still ongoing, privatisation of the entire industry was started and completed
between 1989 and 1996.5
We use aggregate data at firm level provided by Associa¸c˜ao Portuguesa de Bancos.
This is an unexplored rich data set that, in addition to the standard financial information,
contains a number of workforce attributes. In particular, besides information on average
wage and total employment at firm level, we know the composition of the workforce in
terms of job hierarchy, type of activity and seniority. We also have information on
cap-ital formation, profitability and productivity measures that are related to labour market
aspects.
Our analysis produces three main findings. First, the restructuring process during
the privatisation period is highly heterogeneous with different firms adopting adjustment
processes with different speeds and intensities. Second, the variable with the largest
contribution to the overall change in the industry is the pay level of workers. Third, we
show that the wage increase during the privatisation period is associated with a substantial
5
In addition to the more general privatisation literature mentioned above, there is also a separate recent literature on privatisation in banking. The Journal of Banking and Finance ran a special issue on the topic in 2005 (Volume 19, Issues 8-9), where main lessons from a wide range of countries are summarised by Megginson (2005) and Clarke et al. (2005). In line with the more general literature, privatisation in banking is generally (but not always) found to improve bank performance, particularly in terms of profitability. Studies of privatisation in banking that incorporates labour market effects are very scarce, an exception being the aforementioned study by Monteiro (2010).
change in the composition, but not the size, of the workforce.
The rest of the paper is organised as follows. In the next section we describe the
institutional setting and the data used in our analysis. In Section 3 we give a brief,
non-technical, description of the Statis methodology. The technical details are relegated to the
Appendix. We present and discuss our empirical results in Section 4 before closing the
paper with a few concluding remarks in Section 5.
2
Institutional background and data
The privatisation program was introduced in the banking sector as a further step in the
successful reform of the Portuguese financial system (OECD, 1999). This reform started
in 1983-84 with the introduction of laws that removed entry barriers while the interest
rates and credit ceilings were gradually liberalised. Privatisation was only implemented
when most of the price regulations were lifted. The first privatisation law was introduced
in 1988 and only allowed partial privatisation of public firms, with the state retaining
51 percent of the equity. For this first phase of privatisation, the government selected
four profitable firms. In April 1990, the law that regulates privatisation, lei Quadro das
Privatiza¸c˜oes, was passed allowing full privatisation of enterprises nationalised after 1974.
The privatisation program was assumed to be an important mechanism for modernising,
improving the performance and increasing the competitiveness of public economic units.
In addition, it was a stated objective to widen the participation of Portuguese citizens in
the ownership of enterprises, particularly among workers and small shareholders.6
The firms subject to privatisation were first transformed into corporations, with a
prior evaluation being made by two independent entities. However, in contrast with the
privatisation program in some other sectors (e.g., electricity and telecommunications),
the government adopted a hands-off policy in the pre-privatisation period (Naumann,
1995, and Sousa and Cruz, 1995), leaving the economic restructuring for future private
owners. The selection of firms for the partial privatisation phase was made based on
performance indicators (OECD, 1989). However, there was no predetermined schedule for
subsequent privatisations (OECD, 1991). Instead, the timetable was largely determined
by the economic and political domestic cycles, and by the international context.
By 1997, 10 out of 12 public banks became fully private: 2 were privatised in 1991,
3 in 1994, and each of the 5 remaining banks were privatised in 1989, 1990, 1992, 1993
and 1996, respectively.7 Privatisations were mainly implemented by public offer, and in
many cases ownership of the privatised banks returned to Portuguese groups who owned
the banks prior to the nationalisation wave in 1974.8 Due to this private-public-private
ownership path, privatisation in Portugal is often referred to as reprivatisation.
In the empirical analysis, as we have mentioned before, we use aggregate data at
firm level provided by Associa¸c˜ao Portuguesa de Bancos. Given the restrictions imposed
by the Statis methodology (see Section 3), we ended up with a balanced panel covering
10 privatised banks observed during 9 years (between 1989 and 1997), with information
on 14 variables that describe both the labour and the product markets of the banking
sector. Table 1 provides summary statistics (means and standard deviations in brackets)
for all variables in the first and last years – 1989 and 1997 – and for the entire period. In
addition to the wage variable (obtained as the ratio of labour costs and total employment),
we also construct 4 product market variables: labour productivity (total sales divided by
employment), capital-labour ratio (total assets divided by employment), market share
(in terms of sales) and profits per employee (total sales net of worker costs divided by
employment).9 All monetary measures are expressed in 1997 real prices. We use the
Consumer Price Index for converting wages per worker and the GDP deflator for the
remaining variables (profits, market share and capital intensity).
Table 1 shows that privatisation in the Portuguese banking industry brought
consider-able changes in both labour and product markets. Between 1989 and 1997 banks became
smaller. Each bank reduced, on average, by 800 workers, which correspond to an
aver-7According to the privatisation literature, the date of the first tranche sale of each firm is considered the date of effective privatisation.
8International investors could buy a limited share of the equity, ranging from 2 to 40 percent of sales. 9
This particular definition of firm profitability is widely used in the rent sharing literature (see, e.g., Hildreth and Oswald, 1997).
age reduction of 100 workers per year. This downsizing implied substantial changes in
the composition of the workforce. Managers and, in particular, workers in technical and
specific occupations became relatively more abundant. On the other hand, the share of
workers in administrative occupations reduced considerably (about 10 percentage points),
although their share in the total workforce is still very large. The restructuring process
also led to changes in terms of tenure of the workforce. The share of workers with tenure
between 6 and 11 years dropped by 16 percentage points in favour of workers with less
or more tenure. In addition, privatisation restructuring meant that banks reinforced the
already major share of workers in commercial activities.
These compositional changes of the workforce were accompanied by an extraordinary
rise in the average wage which almost tripled between 1989 and 1997. During this period,
there was a general increase in the wage level in virtually all economic sectors,
correspond-ing to the fast economic growth in Portugal after its membership in the European Union in
1986. Nevertheless, Monteiro (2009) finds that earnings in banking rose significantly also
in relative terms, a result that is robust to the use of different control groups. Thus, this
rise in wages reflects particularly the effects of the reforms (liberalisation and privatisation)
in the banking sector which led to considerable improvements in terms of the performance
of banks. For instance, the OECD (1999) survey reports a continuous increase in the
productivity level (implying a reduction in operating/staff costs from 1.53 percent of
av-erage assets in 1991 to 0.98 percent in 1997) and also a remarkable improvement in the
profitability rate (return to equity) after 1995. This global improvement in the efficiency
level of the industry is also confirmed by Pinho (1999), who finds that the performance
improvements are more pronounced among the privatised banks. Even though the
priva-tised banks did not increase, on average, their market shares, the figures in Table 1 clearly
show that labour productivity, capital per worker and profitability increased remarkably
Table 1: Summary statistics by year.
Variables 1989 1997 All
Employment per firm 3816 2992 3454
(2173.9) (1459.2) (1768.3) Share of workers by occupation
Managerial 0.150 0.209 0.175 (0.027) (0.030) (0.037) Technical 0.117 0.168 0.143 (0.042) (0.055) (0.059) Administrative 0.733 0.623 0.682 (0.034) (0.065) (0.072) Share of workers by main activity
Commercial 0.521 0.595 0.564 (0.094) (0.097) (0.098) Other 0.479 0.405 0.436 (0.094) (0.097) (0.098) Share of workers with tenure
Below 6 years 0.086 0.141 0.136 (0.062) (0.104) (0.101) Between 6 and 11 years 0.282 0.119 0.141 (0.120) (0.076) (0.106) Greater than 11 years 0.631 0.740 0.723 (0.150) (0.141) (0.139) Wage (Prices=1997) 19.771 58.633 34.910 (1.694) (57.491) (28.558) Labour productivity (Prices=1997) 187554 459517 316438 (141802.2) (331809.6) (238036.7) Capital labour ratio (Prices=1997) 6797 49907 23310 (8778.0) (77768.0) (35876.0)
Market share 0.058 0.047 0.051
(0.031) (0.036) (0.034) Profit per worker (Prices=1997) 13492 63591 36979 (16663.7) (50386.4) (33856.9)
3
The Statis methodology
The Statis methodology is a three-way exploratory technique based on Principal
Compo-nents Analysis (PCA), commonly used in multivariate data analysis, that enables us to
analyse several data tables of individuals (in our case: banks) described by quantitative
variables, evaluated in different moments in time or circumstances. This methodology
includes two methods: the Statis method, which focuses on the distances between
individ-uals and requires that individindivid-uals are the same in all data tables (although the variables
may differ), and the dual Statis method, that focuses on the relations between variables
and requires that variables are the same in all data tables (although the individuals may
differ).
The Statis (dual Statis) method involves three steps. In the first step, termed
inter-structure, we define an object representative of each data table that describes the structure
of the individuals (variables) in the table. The series of data tables are then globally
com-pared. In the second step, termed intrastructure, we define a common structure of the
individuals (variables) in all the tables, called the compromise, that summarizes the
in-formation included in all data tables. In the last step of the method, we identify the
individuals (variables) responsible for the deviations between the series of data tables,
and we determine which individuals (variables) contribute the most (or least) to the
ob-served differences among the series of tables. In the Appendix we offer a more elaborate
and technical description of these methods. See also L’Hermier des Plantes (1976), Lavit
(1988) and Lavit et al. (1994) for further details.
4
Results
We use the Statis methodology to explore two different issues regarding the restructuring
process of the banking sector. The first issue is to identify a pattern of total changes –
across years and across individual banks. Juxtaposing this information with bank-specific
restructuring process. We will refer to this as the dynamics of restructuring. The second
issue is to decompose the total changes observed and identify which variables made the
largest contributions to the restructuring process. We will refer to this as the content of
restructuring.
4.1 Dynamics of restructuring
We start by identifying the points in time (i.e., years) in which the banks restructure
more (less). For this purpose, we compute the matrix of RV coefficients, which allows to
measure the level of similarity (or the global association) between two data matrices.10 In
our setting, the RV coefficient measures the level of similarity, and thus distance, between
two objects that summarise the structure of banks in any two points in time. The RV
coefficient varies between 0 and 1, where a higher value means a higher closeness between
the two objects being compared. The magnitude of the RV coefficient depends on the
number of individuals (banks), on the number of variables and on the dimensionality
(the covariance structure) of each data set (Josse et al., 2008). Since the individuals
and the variables are constant throughout our analysis, this means that changes in the
RV coefficient can be attributed exclusively to changes in the dimensionality of the data
set. Thus, we can interpret a higher (lower) RV coefficient as a lower (higher) level of
restructuring having taken place.
In terms of computation, due to the enormous difference in the magnitude of variables
from Table 1, the RV coefficients shown in Table 2 are obtained after normalising the
referred variables. Regarding the variables that characterise the employment structure (in
terms of occupation, main activity and seniority) and which together sum to one, we follow
the econometrics tradition of selecting only two variables describing the occupational and
seniority categories and one variable describing the main activity of workers.11 In the end
we use 11 variables: 7 labour and 4 product market variables.12
The RV coefficient compares globally the structure of banks between the years defined
10
Initially Escoufier (1973) introduced the RV coefficient as a measure of similarity between squared symmetric matrices. If we need to compare rectangular matrices, as it happens in our setting, the first
Table 2: Matrix of RV coefficients. 89 90 91 92 93 94 95 96 97 89 1.000 90 0.945 1.000 91 0.953 0.969 1.000 92 0.937 0.926 0.953 1.000 93 0.883 0.858 0.859 0.931 1.000 94 0.899 0.861 0.859 0.931 0.965 1.000 95 0.835 0.857 0.804 0.845 0.838 0.882 1.000 96 0.828 0.861 0.805 0.864 0.872 0.892 0.981 1.000 97 0.776 0.828 0.773 0.844 0.850 0.837 0.913 0.940 1.000
by the first row and column. For instance, focussing on the first column, we observe that
the RV coefficient is gradually reduced (except for 1994), implying that, compared to
1989, banks become more dissimilar, and thus restructure more, over time. In terms of
dynamics, we focus on the figures immediately below the diagonal, which compare the
level of similarity/restructuring between each year and the preceding year. Table 2 shows
that the RV coefficient is high in any pair of consecutive years. This suggests a
contin-uous adjustment to the privatisation reform. Nevertheless, we can identify two ‘peaks’
in the restructuring process. The largest differences in the RV coefficients, compared to
preceding years, are observed in 1993 and 1995. Interestingly, this is consistent with the
aforementioned findings reported by OECD (1999), referring to the year 1995 as a
turn-ing point in the performance of the bankturn-ing sector. In contrast, in 1991 and 1996 we
observe the strongest similarities between two pairs of consecutive years. Therefore, the
graphical representation of the centered interstructure (Figure 1), where the axes of plane
1-2 explain 71.4% of the total variance, allows us to identify three distinct periods of the
restructuring process.13 The three periods are 1989-1992, 1993-1994 and 1995-1997, and
they are identified by the fact that total changes across banks are larger between than
step is to transform them into square matrices. The exact transformation is described in the Appendix. 11
Results do not qualitatively depend on the exclusion or inclusion of these three variables.
12We classify variables that include information about sales as product market variables (e.g., labour productivity).
13
Figure 1 is a two-dimensional representation of the degree of (dis)similarity across years, where a larger distance between two points (years) indicate a stronger dissimilarity between these two years.
within these periods. ● ● ● ● ● ● ● ● ● −0.3 −0.2 −0.1 0.0 0.1 0.2 0.3 −0.3 −0.2 −0.1 0.0 0.1 0.2 1st axis 2nd axis 89 90 91 92 93 94 95 96 97
Figure 1: Euclidean image of the interstructure (Statis).
Are these changes directly related with the privatisation reform? Are banks
homo-geneous in terms of speed of restructuring? Do banks mainly adjust before or after the
reform? Is the adjustment immediately after (before) the reform or is it delayed for some
time after privatisation? In order to answer these questions, we proceed by identifying
how much of the total changes can be explained by each bank, and relate these individual
contributions to the respective privatisation dates. The decomposition of the squared
dis-tances across the banks allows us to identify which banks experienced the largest changes
and at which points in time. Table 3 indicates for each bank the year of privatisation
(column 1) and the respective contribution to the total variation between two consecutive
years (columns 2 to 9) or between the most diverging years 1989 and 1997 (column 10).
Column 11 shows the contribution of each bank to the total variation considering all years.
Naturally, all columns sum to 100%. In order to facilitate the reading of the table, we
are larger than 10 percent.14 We then interpret the contribution of each bank considering
the global size of the RV coefficient (from Table 2) and the date of privatisation.
Table 3: Decomposition of the distance between two years across banks in percentage.
Banks 90 91 92 93 94 95 96 97 89-97 Mean (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) BTA 89 25.6 8.7 8.1 10.1 31.8 4.7 12.9 14.5 12.0 8.3 BPA 90 7.9 5.2 14.4 16.4 17.6 8.8 16.7 14.5 7.3 9.4 BES 91 3.3 5.5 3.0 15.6 0.6 2.8 6.0 7.0 4.6 5.5 BFB 91 2.8 9.5 5.7 8.2 9.8 4.9 6.4 2.0 6.9 6.8 CPP 92 4.6 17.3 4.1 7.3 4.6 2.7 8.5 9.1 5.7 5.6 UBP 93 13.5 6.4 2.9 4.1 13.8 7.2 7.4 11.3 5.8 6.3 BPSM 94 21.0 14.0 6.5 4.2 2.3 40.2 3.4 6.2 15.3 16.8 BFN 94 8.1 15.2 15.3 9.4 6.0 20.8 24.8 20.7 16.8 15.1 BBI 94 7.4 1.8 4.8 2.2 2.1 3.7 10.6 5.5 5.9 4.4 BCA 96 5.8 16.4 35.1 22.6 11.4 4.2 3.3 9.2 19.6 22.0
Some interesting conclusions can be reached. First, it is difficult to establish a clear
relationship between the date of privatisation and the starting point (in time) of
restruc-turing. There is a clear pattern for some banks, such as BTA, BPSM and BFN. These
three banks are responsible for a considerable fraction of all changes that occurred in the
year immediately after their respective dates of privatisation. Moreover, for BPSM and
BFN, the contribution reached 61% of the largest change that occurred throughout the
period of analysis. For other banks, such as BCA, the major adjustments occurred a
considerable amount of time before the reform was implemented (3 to 5 years). For yet
other banks, such as BPA and BES, restructuring took more clearly place 2 to 3 years
after the reform was implemented. For the remaining banks, in particular BBI and BFB,
it is not possible to detect any discernible relationship between the reform and the
ex-tent of restructuring. Nevertheless, regarding the banks for which it is possible to relate
restructuring to the privatisation date, the results appear to suggest different speeds of
adjustment. Some banks, such as BTA and BPSM, seem to have adjusted rapidly (one
year after the ownership change), while BFN and BPA seem to have adopted a more
prolonged and even delayed adjustment process.
In addition, banks did not contribute equally to the total restructuring process. When
we compare the most diverging years – 1989 and 1997 – 4 out of 10 banks explain more
than 60% of the total changes between these two years. Moreover, if we consider all years,
a similar picture is painted: 3 out of the 4 aforementioned banks explain a large proportion
(53.9%) of all changes between any two years. This finding suggests that the privatised
banks constituted a heterogeneous group before the implementation of the reform and
therefore adjusted at different timing, speed and intensity. The same conclusion can be
reached by inspecting Figures 2 and 3, which represent the trajectory of each bank around
its compromise position in planes 1-2 and 1-3 (the first three axes explain 78.5% of the
total variance).15
For each bank, each dot represents one year (with the years 1989 and 1997 being
ex-plicitly indicated). The bold and big dots represent the compromise (average) position in
the period of analysis. The first axis is positively correlated with the variables profits per
worker, labour productivity and share of workers in technical occupations while negatively
correlated with the variable share of workers in commercial activities.16 The second axis
is positively correlated the variables bank employment, market share and labour
productiv-ity.17 Finally, the third axis is positively correlated with labour productivity and negatively
correlated with the share of workers in managerial occupations.18 The banks that exhibit
a larger trajectory or movement across both axes in the two planes correspond to the
banks that restructured more. A narrower trajectory shows a lower level of adjustment.
15
We show the trajectory of each bank in three planes in order to maximise the share of total variance explained. The first two axes explain 68.4% of the total variance, while 78.5% is explained by the addition of the third axis.
16
A bank like BFN, that has high profits and labour productivity and a high share of technical workers appears on the positive side of the axis, while banks like BPSM and BBI, that have a large share of workers in commercial activities appear on the negative side of the first axis.
17Banks like BPA, BES and BPSM, that are recorded with high values of these three variables appear on the positive side of the second axis.
18Banks with high (low) figures on labour productivity and low (high) figures on the share of workers in managerial occupations, such as BPSM and BCA (BTA and BFB), appear on the positive (negative) side of the third axis.
● ● ● ● ●● ●● ● −0.2 0.0 0.2 −0.4 0.0 BTA 1st axis 2nd axis ●89 97 ● ● ● ●● ● ● ● ● −0.2 0.0 0.2 −0.4 0.0 BPA 1st axis 2nd axis ● 89 97 ● ● ● ● ●● ●● ● −0.2 0.0 0.2 −0.4 0.0 BES 1st axis 2nd axis ● 8997 ●● ● ● ● ● ● ● ● −0.2 0.0 0.2 −0.4 0.0 BFB 1st axis 2nd axis ●89 97 ● ● ● ● ● ● ● ● ● −0.2 0.0 0.2 −0.4 0.0 CPP 1st axis 2nd axis ● 89 97 ●● ● ● ● ● ● ● ● −0.2 0.0 0.2 −0.4 0.0 UBP 1st axis 2nd axis ● 89 97 ● ● ● ● ● ● ● ● ● −0.3 −0.1 0.1 −0.4 0.0 BPSM 1st axis 2nd axis ● 89 97 ● ● ● ● ● ● ● ● ● 0.5 0.7 0.9 −0.4 0.0 BFN 1st axis 2nd axis ● 89 97 ● ● ● ● ● ● ● ● ● −0.4 −0.2 0.0 −0.4 0.0 BBI 1st axis 2nd axis ●89 97 ● ● ● ● ● ● ● ● ● −0.2 0.0 0.2 −0.4 0.0 BCA 1st axis 2nd axis ● 89 97
Figure 2: Trajectory of each bank around its compromise position (plane 1-2).
4.2 Content of restructuring
A further decomposition of the restructuring process allows us to explore which variables
contribute more to the overall magnitude of restructuring. Table 4 replicates Table 3
by decomposing the squared distances between the correlation matrices across years
● ● ● ● ● ● ● ● ● −0.2 0.0 0.2 −0.15 0.05 0.20 BTA 1st axis 3rd axis ●89 97 ● ●● ● ● ● ● ● ● −0.2 0.0 0.2 −0.15 0.05 0.20 BPA 1st axis 3rd axis ● 8997 ● ● ● ● ● ● ● ● ● −0.2 0.0 0.2 −0.15 0.05 0.20 BES 1st axis 3rd axis ● 8997 ● ● ● ● ● ● ● ● ● −0.2 0.0 0.2 −0.15 0.05 0.20 BFB 1st axis 3rd axis ● 89 97 ● ● ● ● ● ● ● ● ● −0.2 0.0 0.2 −0.15 0.05 0.20 CPP 1st axis 3rd axis ● 89 97 ● ● ● ● ● ● ● ● ● −0.2 0.0 0.2 −0.15 0.05 0.20 UBP 1st axis 3rd axis ● 89 97 ● ● ● ● ● ● ● ● ● −0.3 −0.1 0.1 −0.15 0.05 0.20 BPSM 1st axis 3rd axis ● 89 97 ● ● ● ● ● ● ● ● ● 0.5 0.7 0.9 −0.15 0.05 0.20 BFN 1st axis 3rd axis ● 89 97 ● ● ● ● ● ● ● ● ● −0.4 −0.2 0.0 −0.15 0.05 0.20 BBI 1st axis 3rd axis ● 89 97 ● ● ● ● ● ● ● ● ● −0.2 0.0 0.2 −0.15 0.05 0.20 BCA 1st axis 3rd axis ● 89
Figure 3: Trajectory of each bank around its compromise position (plane 1-3).
figures variables whose contribution to the overall changes is above 9%, which corresponds
approximately to the division of 100% by 11 variables.
Considering all years (column 11), the variables that explain most of the restructuring
are wage per worker, share of workers with tenure between 6 and 11, capital labour ratio
and workers in managerial occupations. Notice that this type of information cannot be
Table 4: Decomposition of the distance between two years across variables in percentage.
Variables 90 91 92 93 94 95 96 97 89-97 Mean (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Employment per firm 6.0 5.1 3.5 7.8 3.2 3.9 7.6 9.3 3.4 Share of workers by occupation
Managerial 18.5 12.8 20.4 7.6 9.2 18.4 16.0 9.3 10.5 10.4 Technical 7.5 5.7 8.3 4.7 6.0 14.9 5.4 8.2 11.8 6.5 Share of workers by main activity
Commercial 4.4 8.0 8.3 7.6 3.4 8.4 10.4 7.4 7.0 5.8 Share of workers by tenure
Below 6 years 8.8 8.5 3.2 7.1 1.4 4.7 8.2 11.2 6.1 6.7 Between 6 and 11 years 3.0 12.3 4.9 7.2 37.6 2.4 30.9 15.8 14.4 12.5 Wage 30.4 16.8 4.2 32.0 5.1 10.2 3.1 8.1 10.5 17.7 Labour productivity 4.1 6.5 9.0 3.0 7.9 12.2 3.5 4.7 10.4 8.4 Capital labour ratio 9.0 12.6 24.5 12.0 6.6 2.9 9.9 11.6 8.5 10.7 Market share 3.4 7.6 6.2 8.0 12.5 13.5 2.0 9.0 8.2 8.9 Profit per worker 5.1 4.1 7.7 2.9 7.0 8.4 3.1 5.4 9.3 7.1
the banking industry is by far the variable that changed the most, with a contribution
of 17.7% of the total variation. Interestingly, this variation reflects mainly substantial
changes in the quality, and not quantity, of the workforce, either in terms of seniority or
occupational category. In particular, the correlation matrices19 across years show that
in the period 1993-1994, higher wages are associated with a larger share of workers with
tenure below 6 years, a larger share of workers in technical occupations and a lower share
of workers in commercial activities, while in the subsequent period 1995-1997, higher
wages are associated with a larger share of workers with tenure below 6 years, a lower
share of workers in managerial occupations and a lower share of workers in commercial
activities. The magnitude of pay level changes might also to reflect gains in terms of
labour productivity, profits and capital per worker, in particular after 1992. In fact, the
correlations between wages, profits, labour productivity and capital per worker become
substantially stronger and positive over the period 1993-1997. This finding suggest that
wages might be responding to the increased profitability of the banks, possibly explained
by more rent sharing in privatised banks. This result is consistent with findings of Monteiro
et al. (2010) for the whole Portuguese economy.
●●● ● ● ● ● ● ● −1.0 0.0 −1.0 0.0 1.0 Employment 1st axis 2nd axis ● 97 89 ● ● ● ● ● ● ● ● ● −1.0 0.0 −1.0 0.0 1.0 Managerial 1st axis 2nd axis 97 89● ● ● ● ● ●● ● ● ● −0.5 0.5 −1.0 0.0 1.0 Technical 1st axis 2nd axis ● 97 89 ● ●● ● ● ● ● ● ● −1.5 −0.5 −1.0 0.0 1.0 Commercial 1st axis 2nd axis ●97 89 ● ● ● ● ● ● ● ● ● 0.0 1.0 −1.0 0.0 1.0 Tenure <6 1st axis 2nd axis ● 97 89 ● ● ● ● ● ● ● ● ● −0.5 0.5 −1.0 0.0 1.0 Tenure 6−11 1st axis 2nd axis ● 97 89 ● ● ● ● ● ● ● ● ● 0.0 1.0 −1.0 0.0 1.0 Wage 1st axis 2nd axis ● 97 89 ● ●● ●● ● ● ●● 0.0 1.0 −1.0 0.0 1.0 Labour productivity 1st axis 2nd axis ● 97 89 ● ●● ● ● ● ● ● ● 0.0 1.0 −1.0 0.0 1.0
Capital labour ratio
1st axis 2nd axis ● 97 89 ●●● ●● ● ● ● ● −0.5 0.5 −1.0 0.0 1.0 Market share 1st axis 2nd axis ● 97 89 ● ●● ● ● ● ●● ● 0.0 1.0 −1.0 0.0 1.0
Profit per worker
1st axis
2nd axis
●
97
Figure 4: Trajectory of each variable around its compromise position (plane 1-2).
Table 4 also suggests that the initial phase of the banks’ restructuring process consisted
mainly of changes in the workforce composition and investment in capital equipment.
● ● ● ● ● ● ● ● ● −1.0 0.0 −1.0 0.0 1.0 Employment 1st axis 3rd axis ● 97 89 ● ● ● ● ● ● ● ● ● −1.0 0.0 −1.0 0.0 1.0 Managerial 1st axis 3rd axis ● 97 89 ● ● ● ● ● ● ● ● ● 0.0 1.0 −1.0 0.0 1.0 Technical 1st axis 3rd axis ● 97 89 ● ● ● ● ● ● ● ● ● −1.5 −0.5 −1.0 0.0 1.0 Commercial 1st axis 3rd axis ● 97 89 ● ● ● ● ● ● ● ● ● 0.0 1.0 −1.0 0.0 1.0 Tenure <6 1st axis 3rd axis ● 97 89 ● ● ●● ● ● ● ● ● −0.5 0.5 −1.0 0.0 1.0 Tenure 6−11 1st axis 3rd axis ● 97 89 ● ● ● ● ● ● ● ● ● 0.0 1.0 −1.0 0.0 1.0 Wage 1st axis 3rd axis ● 97 89 ● ● ● ● ● ● ●●● 0.0 1.0 −1.0 0.0 1.0 Labour productivity 1st axis 3rd axis ● 97 89 ● ● ● ● ● ● ● ● ● 0.0 1.0 −1.0 0.0 1.0
Capital labour ratio
1st axis 3rd axis ●97 89 ● ● ● ● ● ● ● ● ● −0.5 0.5 −1.0 0.0 1.0 Market share 1st axis 3rd axis ● 97 89 ● ● ● ● ● ● ●● ● 0.0 1.0 −1.0 0.0 1.0
Profit per worker
1st axis
3rd axis
●
97
89
Figure 5: Trajectory of each variable around its compromise position (plane 1-3).
market share and labour productivity – which fed further changes in wages and in
em-ployment structure. As before, it is possible to inspect visually which variables changed
the most. The trajectory of each variable around its compromise position in planes 1-2
and 1-3 is presented in Figures 4 to 5. The three axes explain 71.7% of total variance.20
20The first axis represents correctly the variables profits per worker, labour productivity, share of workers with tenure below 6 years, wage per worker, capital labour ratio, share of workers in technical occupations and share of workers in commercial activities. Together, these 7 variables explain 96% of the first axis.
In accordance with Table 4, the variables wage per worker, managerial occupation, tenure
between 6 and 11 years and capital labor ratio exhibit ample trajectories across one or
both axes, implying sizeable changes in the variables over time.
5
Concluding remarks
This paper examines the restructuring process of 10 privatised banks during a 9-year
pe-riod, using the Statis and the dual Statis approach. Our empirical findings pinpoint some
important lessons. First, the analysis of privatisation effects using data aggregated at
in-dustry level can potentially be a dangerous exercise as it might obscure diverse adjustment
processes occurring at individual firm level. Furthermore, apart from identifying wages as
the most important variable in the restructuring process, our results also indicate
signifi-cant skill compositional effects of privatisation, a dimension almost absent in the literature
concerning the labour market effects of privatisation. Finally, we also provide empirical
evidence suggesting that privatisation is associated with a higher level of rent sharing, a
topic that deserves further research.
6
Appendix: Statis and dual Statis
Let (Xk)n×pk, k = 1, . . . , K, be the data table associated with the kth point in time or
circumstance, where n refers to the total number of individuals and pk is the number of
variables in the kth data table.
The Statis method: Let Qk be the metric in the individuals space (in general defined
by the identity matrix or by a diagonal matrix whose main elements are the
recipro-cal of the variance of variables) which enables us to determine the distance between
two individuals and let D be the metric in the variables space (in general defined by
a diagonal matrix whose elements are the weights associated to the individuals) which
The second axis is explained mainly (81% ) by the variables market share, bank employment, workers in managerial occupations and capital labour ratio. The variables share of workers with tenure between 6 and 11, share of workers in managerial occupations and capital labour ratio explain 76% of the third axis.
enables us to determine the correlation between two variables. In the interstructure
step we associate to each Xk a matrix Wk of the scalar products between
individu-als, the object representative of each data table, given by Wk = XkQkXkT, where XkT
denotes the transpose matrix of Xk. The distance between objects at stages k and
k0 is given by the scalar product of Hilbert-Schmidt, hWk, Wk0i
HS = T r (WkDWk0D) , where T r denotes the trace operator of a matrix, being kWkk = phWk, WkiHS. The vectorial correlation coefficient RV proposed by Robert and Escoufier (1976) is
equiva-lent to the scalar product of Hilbert-Schmidt between normed objects and is defined by
RV (k, k0) = W∗ k, Wk∗0 HS = T r (WkDWk0D) / q T r (WkD)2T r (Wk0D)2, where W∗ i de-notes the ith normed object Wi/ kWik. The RV coefficient varies between 0 and 1, meaning
that the higher it is the closer are the two objects being compared. The distance between
the normed objects is given by dHS Wk∗, Wk∗0 =
Wk∗− Wk∗0
= p2 − 2RV (k, k0). De-noting by S the matrix of coefficients RV and by ∆ the diagonal matrix of weights πk
associated to each table, a principal component analysis based on the matrix S∆ gives
us the euclidean image of the series of data tables. The coordinates of the points Ak
associated with the data tables on the ith axis, are the components of the vector√τiγi,
where τi represents the ith largest eigenvalue of S∆ associated with the eigenvector γi.
Notice that if the weights πk are equal, it is enough to base the PCA on the matrix S.
For obtaining a centered euclidean image of the data tables, we base the PCA on the
matrix ∼
S = IK− 11T∆ S IK− ∆11T , where IK is the identity matrix of order K and 1 is a vector of dimension K with all components equal to 1. In the intrastructure
step, the compromise is the object W , defined by the weighted mean W = K P k=1
αkWk∗,
where the coefficients αk are given by αk = πkγ1k /√τ1 and γ1k is the kth coordinate of
the vector γ1. A PCA based on the matrix W enables us to obtain the euclidean image
of the compromise. The coordinates of the points Bi, i = 1, . . . , n, associated with the
individuals on the kth axis of the euclidean image of the compromise are the components
of the vector√µkεk, where µk denotes the eigenvalue of the matrix W D associated with
to interpret the compromise axes and the compromise positions of the individuals. In the
last step of the method, we identify the individuals responsible for the deviations between
the series of data tables, through the decomposition of the squared distances between
two pairs of objects into percentages of individuals’ contributions, i.e., we calculate the
quantities Cind i,d2
HS = (dii/d 2 HS(Wk, Wk0)) n P j=1 djj h Wkij − Wkij0 i2
, where Cind i,d2
HS
repre-sents the contribution of the ith individual to the squared distance d2HS, dii denotes the
ith diagonal element of the matrix D and Wkij denotes the ij-element of the matrix Wk.
For visualizing graphically the individuals responsible for the deviations between the series
of data tables, we represent the different positions of the individuals for each object on
the compromise euclidean image, i.e., their trajectories. The coordinates of the points
B1k, ..., Bnk , k = 1, ..., K on ith axis are given by WkDεi/ √
µi.
Dual Statis method: Let Q and Dk be the metrics in the individuals and variables space,
respectively, defined as in the Statis method. The object representative of each data table is
defined by either the covariance or correlation matrix (in case of standardized data), given
by Vk = XkTDkXk. The scalar product of Hilbert-Schmidt between two objects at stages k
and k0 is defined by hVk, Vk0i
HS = T r (QVkQVk0) . The diagonalization of the matrix Z∆,
where Z denotes the matrix of the scalar products between the objects, allows us to obtain
the euclidean image of the series of data tables. The compromise is the object given by V = K
P k=1
βkVk. The diagonalization of the matrix V Q enables us to obtain the euclidean image of
the compromise. Finally, we decompose the squared distances between two pair of objects
into percentages of variables’ contributions, i.e., we calculate the quantities Cvar i,d2
HS = (qii/d2HS(Vk, Vk0)) p P j=1 qjj h Vkij − Vkij0 i2
, where Cvar i,d2
HS represents the contribution of the
ith variable to the squared distance d2HS, qiidenotes the ith diagonal element of the matrix
Q and Vkij denotes the ij-element of the matrix Vk. We also represent the variables’
trajectories on the compromise euclidean image.
References
[2] Brainerd, E., 1998. Winners and losers in Russia’s economic transition. American
Economic Review, 88, 1094–1116.
[3] Brainerd, E., 2002. Five years after: the impact of mass privatisation on wages in
Russia, 1993-1998. Journal of Comparative Economics, 30, 160–190.
[4] Brown, J., Earle, J., Vakhitov, V., 2006. Wages, layoffs, and privatisation: evidence
from Ukraine. Journal of Comparative Economics, 34, 272–294.
[5] Brown, D., Earle, J., Telegdy, J. ´A., 2010. Employment and wage effects of
privati-sation: evidence form Hungary, Romania, Russia and Ukraine. Economic Journal,
forthcoming.
[6] Clarke, G., Cull, R., Shirley, M., 2005. Bank privatisation in developing countries: a
summary of lessons and findings. Journal of Banking and Finance, 29, 1905–1930.
[7] Chong, A., Leon, G., 2009. Privatised firms and labour outcomes in emerging markets.
Journal of Development Studies, 45, 1513–1525.
[8] Djankov, S., Murrel, P., 2002. Enterprise restructuring in transition: a quantitative
survey. Journal of Economic Literature, 40, 739–792.
[9] Escoufier, Y., 1973. Le traitement des variables vectorielles. Biometrics, 29, 751–760.
[10] Estrin, S., Hanousek, J., Kocenda, E., Svejnar, J., 2009. The effects of privatisation
and ownership in transition economies. Journal of Economic Literature, 47, 699–728.
[11] Haskel , J., Szymansky, S., 1993. Privatization, liberalization, wages and employment:
theory and evidence for the UK. Economica, 60, 161–181.
[12] Hildreth, A.K., Oswald, A.J., 1997. Rent-sharing and wages: evidence from company
and establishment panels. Journal of Labor Economics, 15, 318–37.
[13] Ho, S., Dong, X., Bowles, P., MacPhail, F., 2002. Privatisation and enterprise wage
structures during transition: evidence from rural industry in China. Economics of
[14] J. Josse, Pag`es J., Husson, F., 2008. Testing the significance of the RV coefficient.
Computational Statistics & Data Analysis, 53, 82–91.
[15] La Porta, R., L´opez-de-Silanes, F., 1999. The benefits of privatisation: evidence from
Mexico. Quarterly Journal of Economics, 114, 1193–1242.
[16] Lavit, C., 1988. Analyse conjointe de tableaux quantitatifs. Editions Masson.
[17] Lavit, C., Escoufier, Y., Sabatier, R., Traissac, P., 1994. The ACT (Statis method).
Computational Statistics and Data Analysis, 18, 97–119.
[18] L’Hermier des Plantes, H., 1976. Structuration des tableaux `a trois indices de la
statistique. Th`ese de 3 `eme cycle, Universit´e de Montpellier.
[19] Liu, P.-W., Xin, M., Zhang, J., 2000. Sectoral gender wage differentials and
discrim-ination in the transitional Chinese economy. Journal of Population Economics, 13,
331–352.
[20] Megginson, W., 2005. The economics of bank privatisation. Journal of Banking and
Finance, 29, 1931–1980.
[21] Megginson, W., Netter, J., 2001. From state to market: a survey of empirical studies
on privatisation. Journal of Economic Literature, 39, 321–389.
[22] Megginson, W., Sutter, N., 2006. Privatisation in developing countries. Corporate
Governance, 14, 234–265.
[23] Monteiro, N. P., 2009. Regulatory reforms and labour earnings in Portuguese banking.
Empirical Economics, 36, 557-574.
[24] Monteiro, N. P., 2010. Using propensity matching estimators to examine the effect of
privatisation on wages. Applied Economics, 42, 1293–1313.
[25] Monteiro, N. P., Portela, M., Straume, O. R., 2010. Firm ownership and rent-sharing.
[26] Munich, D., Svejnar, J., Terrel, K., 2005. Returns to human capital under the
commu-nist wage grid and during the transition to a market economy. Review of Economics
Statistics, 87, 100–123.
[27] Naumann, R., 1995. Privatiza¸c˜oes e reestrutura¸c˜oes. O desafio para o movimento
sindical em Portugal, Funda¸c˜ao Friedrich Ebert, Lisboa.
[28] OECD, 1989. OECD Economic Surveys, Portugal, 1988-89, OECD, Paris.
[29] OECD, 1991. OECD Economic Surveys, Portugal, 1990-91, OECD, Paris.
[30] OECD, 1999. OECD Economic Surveys, Portugal, 1998-99, OECD, Paris.
[31] Parker, D., Martin, S., 1996. The impact of UK privatisation on employment, profits
and the distribution of business income. Public Money & Management, 16, 31–37.
[32] Peoples, J, Talley, W., 2001. Black-white earnings differentials: privatisation versus
deregulation. American Economic Review, 91, 164–168.
[33] Pinho, P. 1999. Reprivatiza¸c˜oes e eficiˆencia no sistema banc´ario Portuguˆes.
Docu-mento de trabalho no 13 (Lisboa: Direc¸c˜ao Geral de Estudos e Previs˜ao, Minist´erio
das Finan¸cas)
[34] Robert, P., Escoufier, Y., 1976. A unifying tool for linear multivariate statistical
methods: the RV coefficient. Applied Statistics, 25, 257–265.
[35] Sousa, F. and Cruz, R., 1995. O Processo de Privatiza¸c˜oes em Portugal, Associa¸c˜ao
Industrial Portuense, Porto.
[36] Villalonga, B., 2000. Privatization and efficiency: differentiating ownership effects
from political, organizational, and dynamic effects. Journal of Economic Behavior
Most Recent Working Paper
NIPE WP 16/2010
Figueiredo, Adelaide, Fernanda Figueiredo, Natália Monteiro e Odd Rune Straume, "Restructuring in privatised firms
:
a Statis approach", 2010NIPE WP 15/2010
Sousa, Ricardo M., “Collateralizable Wealth, Asset Returns, and Systemic Risk: International Evidence", 2010
NIPE WP 14/2010
Sousa, Ricardo M., “How do Consumption and Asset Returns React to Wealth Shocks? Evidence from the U.S. and the U.K", 2010
NIPE WP 13/2010
Monteiro, Natália., Miguel Portela e Odd Rune Straume, "Firm ownership and rent sharing", 2010
NIPE WP 12/2010
Afonso, Oscar, Sara Monteiro e Maria Thompson., "A Growth Model for the Quadruple Helix Innovation Theory ", 2010
NIPE WP 11/2010
Veiga, Linda G.,"Determinants of the assignment of E.U. funds to Portuguese municipalities", 2010
NIPE WP 10/2010
Sousa, Ricardo M., "Time-Varying Expected Returns: Evidence from the U.S. and the U.K", 2010
NIPE WP 9/2010
Sousa, Ricardo M., "The consumption-wealth ratio and asset returns: The Euro Area, the UK and the US", 2010
NIPE WP 8/2010
Bastos, Paulo, e Odd Rune Straume, "Globalization, product differentiation and wage inequality", 2010
NIPE WP 7/2010
Veiga, Linda, e Francisco José Veiga, “Intergovernmental fiscal transfers as pork barrel”, 2010 NIPE WP
6/2010
Rui Nuno Baleiras, “Que mudanças na Política de Coesão para o horizonte 2020?”, 2010 NIPE WP
5/2010
Aisen, Ari, e Francisco José Veiga, “How does political instability affect economic growth?”, 2010
NIPE WP 4/2010
Sá, Carla, Diana Amado Tavares, Elsa Justino, Alberto Amaral, "Higher education (related) choices in Portugal: joint decisions on institution type and leaving home", 2010
NIPE WP 3/2010
Esteves, Rosa-Branca, “Price Discrimination with Private and Imperfect Information ”, 2010
NIPE WP 2/2010
Alexandre, Fernando, Pedro Bação, João Cerejeira e Miguel Portela, “Employment, exchange rates and labour market rigidity”, 2010
NIPE WP 1/2010
Aguiar-Conraria, Luís, Pedro C. Magalhães e Maria Joana Soares,
“On Waves in War and
Elections - Wavelet Analysis of Political Time-Series”, 2010
NIPE WP 27/2009
Mallick, Sushanta K. e Ricardo M. Sousa,
“Monetary Policy and Economic Activity in the
BRICS”, 2009
NIPE WP 26/2009
Sousa, Ricardo M.,
“
What Are The Wealth Effects Of Monetary Policy?”, 2009
NIPE WP 25/2009
Afonso, António., Peter Claeys e Ricardo M. Sousa,
“Fiscal Regime Shifts in Portugal”,
2009
NIPE WP 24/2009
Aidt, Toke S., Francisco José Veiga e Linda Gonçalves Veiga,
“Election Results and
Opportunistic Policies: A New Test of the Rational Political Business Cycle Model”, 2009
NIPE WP 23/2009
Esteves, Rosa Branca e Hélder Vasconcelos,“
Price Discrimination under Customer
Recognition and Mergers
”, 2009NIPE WP 22/2009
Bleaney, Michael e Manuela Francisco,“
What Makes Currencies Volatile? An Empirical
Investigation
”, 2009NIPE WP 21/2009
Brekke, Kurt R. Luigi Siciliani e Odd Rune Straume,“Price and quality in spatial competition”, 2009
NIPE WP 20/2009
Santos, José Freitas e J. Cadima Ribeiro,“Localização das Actividades e sua Dinâmica”, 2009 NIPE WP
19/2009
Peltonen, Tuomas A.,Ricardo M. Sousa e Isabel S.Vansteenkiste“Fundamentals, Financial Factors and The Dynamics of Investment in Emerging Markets”, 2009
